Research map / Label-efficient and robust learning / OOD and anomaly detection
DenseHybrid: Hybrid Anomaly Detection for Dense Open-Set Recognition
Matej Grcić, Petra Bevandić, Siniša Šegvić
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| fuse generative p(x) and discriminative P(d_in|x) into a hybrid score (ablation) | DenseHybrid · Fishyscapes LostAndFound (val) | AP: 60.5 -> 63.8 (generative), 42.9 -> 63.8 (discriminative) | Improved |
| fuse generative and discriminative scores into hybrid (ablation) | DenseHybrid · Fishyscapes Static (val) | AP: 54.2 -> 60.0 (generative), 47.8 -> 60.0 (discriminative) | Improved |
| fuse generative and discriminative scores into hybrid (ablation) | DenseHybrid · StreetHazards | open-world mean IoU: 36.4 -> 45.8 (discriminative) | Improved |
| add discriminative term to generative score (ablation) | DenseHybrid · StreetHazards | AP: 30.0 -> 30.2 | No clear effect |
| DenseHybrid vs ML-based open-set recognition | LDN-121 · StreetHazards | open-world mean IoU: 41.2 -> 45.8 | Improved |
| add known inlier instances (train+val) to training (ablation) | DenseHybrid · Fishyscapes Static | AP: 60.0 -> 88.5 | Improved |
| train on train+val instead of train (ablation) | DenseHybrid · Fishyscapes Static | FPR95: 4.9 -> 1.1 | Improved |
| add DenseHybrid module to LDN-121 | LDN-121 · inference cost, 2 MP images | overhead: 0.1 GFLOPs and 2.8ms | No clear effect |
| increase distance from camera | DenseHybrid · LostAndFound range analysis | AP: 90.7 (5-10 m) -> 82.1 (45-50 m) | Hurt |
| DenseHybrid vs SynBoost on farther ranges | DenseHybrid · LostAndFound range analysis | AP: 53.3 -> 82.1 (45-50 m) | Improved |
| anomaly detector on open-world recognition | best detectors · StreetHazards | mIoU: over 60% closed world -> 46% open world | Hurt |
| add DenseHybrid fine-tuning to the segmentation model | DenseHybrid · Fishyscapes Cityscapes closed world | mIoU: 81.4 vs 81.0 | No clear effect |
Later papers that compared with this method
1 later accepted paper reports DenseHybrid as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Residual Pattern Learning for Pixel-Wise Out-of-Distribution Detection in Semantic Segmentation (ICCV 2023 · better on 1 of 1 shared result)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Dense Projection for Anomaly Detection (AAAI 2024 · similarity 0.89)
- Anomaly Heterogeneity Learning for Open-set Supervised Anomaly Detection (CVPR 2024 · similarity 0.88)
- ResAD: A Simple Framework for Class Generalizable Anomaly Detection (NeurIPS 2024 · similarity 0.87)
- Dinomaly: The Less Is More Philosophy in Multi-Class Unsupervised Anomaly Detection (CVPR 2025 · similarity 0.87)
- Generating and Reweighting Dense Contrastive Patterns for Unsupervised Anomaly Detection (AAAI 2024 · similarity 0.87)
- A Dataset for Semantic Segmentation in the Presence of Unknowns (CVPR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
